Detecting Adverse Events Following Acute Care Encounters among Veterans Prescribed Non-Steroidal Anti-Inflammatory...

Area of Science:

• Pharmacovigilance
• Health Informatics
• Clinical Epidemiology

Background:

• Nonsteroidal anti-inflammatory drugs (NSAIDs) are commonly prescribed but carry risks for adverse events (AEs), including gastrointestinal, renal, and cardiovascular complications.
• These risks are amplified in vulnerable populations such as older adults, patients with chronic kidney disease (CKD), and those on polypharmacy.
• Understanding AE occurrence within specific healthcare systems like the Veterans Health Administration (VHA) is crucial for targeted interventions.

Purpose of the Study:

• To develop and evaluate predictive models for adverse events (AEs) in Veterans prescribed nonsteroidal anti-inflammatory drugs (NSAIDs).
• To identify specific risk factors associated with NSAID-related AEs within the VHA.
• To assess the performance of machine learning models in predicting AEs in this population.

Main Methods:

• Utilized eXtreme Gradient Boosting (XGBoost), LASSO regression, and logistic regression to model AE occurrence.
• Defined AEs as a composite outcome including acute coronary syndrome, acute kidney injury, allergic reactions, gastrointestinal bleeding, or GERD within 2-30 days post-encounter.
• Evaluated model performance using Area Under the Receiver Operating Characteristic Curve (AUC) and calibration plots, calculating observed-to-expected (OE) ratios.

Main Results:

• Analyzed 1,112,819 encounters for 736,677 unique Veterans, identifying 50,278 (4.5%) AEs.
• XGBoost and LASSO models demonstrated strong performance with AUC values between 0.77 and 0.79.
• Gastroesophageal reflux disease (GERD) and advanced CKD (stage 4-5/dialysis) were significantly associated with increased AE risk; OE ratios varied by VHA site.

Conclusions:

• Risk adjustment models exhibited good discrimination and identified significant site-specific variations in NSAID-related AEs.
• These predictive models show potential utility for clinical decision support systems.
• The findings support the development of targeted population health interventions to mitigate NSAID-associated risks.